By: David Schroeder
Melt ponds are puddles of water that form on top of sea ice when the snow and ice melts (see Figure). Not all the water drains immediately into the ocean, but it can stay and accumulate on top of the sea ice for several weeks or months (Ref: https://blogs.reading.ac.uk/weather-and-climate-at-reading/2017/melt-ponds-over-arctic-sea-ice/
Figure: Melt ponds on sea ice (Credit: Don Perovich)
A momentous field campaign was carried out in 1998 on the Arctic sea ice: the Surface Heat Budget of the Arctic Ocean (SHEBA) experiment (https://www.nsf.gov/pubs/2003/nsf03048/nsf03048_3.pdf) – a role model for the latest and largest Arctic expedition MOSAIC in 2019/2020 (https://mosaic-expedition.org/expedition/). One aim was to understand and quantify the sea ice-albedo feedback mechanism on scales ranging from meters to thousands of kilometers. The differences in albedo (fraction of shortwave radiation reflected at the surface and, thus, not used to heat the surface) between snow-covered sea ice (~85%), bare sea ice (~60-70%), ponded sea ice (~30%) and open water (<10%) are huge and cause the most important feedback for sea ice melt: The more and the earlier snow and ice melts, the larger the pond and open water fraction, the more shortwave radiation will be absorbed increasing the melting. Melt ponds play an important part in the observed reduction and thinning of Arctic sea ice during last decades.
Continuous SHEBA measurements over the whole melt season in 1998 allowed the development of models representing the melting cycle: from the onset of melt pond formation, spreading, evolution and drainage over late spring and summer, towards freeze-up in the late summer and autumn. Starting with a one-dimensional heat balance model (Taylor and Feltham, 2004), it took about 10 years to develop a pond model suitable for a Global Climate Model (GCM) (Flocco et al., 2010; 2012). Melt pond formation is controlled by small-scale sea ice topography. This is not available in a GCM with coarser resolution. However, we could use the sub-gridscale ice thickness distribution (5 different ice thickness categories for each grid cell) as a proxy for topography and simulate the evolution of pond fraction assuming melt water runs from the thicker ice to the thinner ice. With further adjustments to the albedo scheme (Ridley et al., 2018), the pond model could finally be used in the UK Climate Model HadGEM3. The HadGEM3 simulations for the latest IPPC report (https://www.ipcc.ch/report/ar6/wg2/) include our pond model.
What is the impact of the melt pond model on the performance of the HadGEM3 simulations? It is noteworthy that HadGEM3 has a stronger climate sensitivity (global warming with respect to CO2 increase) compared to its predecessor HadGEM2 or most other climate models (Mehl et al., 2020). But is this due to the melt ponds? Lots of model components were changed at the same time, so it is impossible to specify the individual impact. To address this, Diamond et al. (2023) carried out HadGEM3 simulations with 3 configurations which only differ with respect to melt pond treatment (our pond scheme, simple albedo tuning to account for the impact of melt ponds and no melt ponds). Historical or future projections would require an ensemble simulation to distinguish between internal variability and impact of pond scheme. Thus, 100 year long constant forcing simulations have been chosen.
While Arctic sea ice results between the simple albedo tuning and our full pond scheme do not differ significantly for pre-industrial conditions, the impact on near future conditions are remarkable: The simple tuning never yields an ice-free summer Arctic, whilst our pond scheme yields an ice-free Arctic 35% of years and raises autumn Arctic air temperatures by 5 to 8 °C. Thus, the pond treatment has a large impact on projections when the Arctic will become ice-free. This is a striking example of the impact